# 6.2.1 Input data from the upstream processing

The input data to the spectroscopic pipeline have been pre-processed by the upstream Initial Data Treatment (IDT) described in Chapter 2, where the spectroscopic telemetry information is reconstructed into several data types (the list of the data types is provided in Section 1.3.4). The data used by the spectroscopic pipeline are: SpectroObservation (Section 2.4.3) containing the three CCD spectra acquired in the transit together with various information related to the observations (like the on-board magnitude, the window size, position and truncation); ObjectLogRvs and ObjectLogRvsRotSmear, containing the detection features including the information on the satellite solar rotation phase; AcShift; PreScan; SpectroObservationVo, containing the spectra of the Virtual Objects (empty windows acquired for calibration purposes).

Also needed are the BaryVeloCorr tables, produced by IDT, containing the barycentric velocity correction to be added to the spectroscopic, Gaia-centric, radial velocity measurements to obtain the velocities relative to the solar system barycentre.

In addition to the IDT data, some other input data are produced by the upstream astrometric processing described in Chapter 3: the AGIS position of the sources, the attitude model OGA3 (Section 3.3.5), the IDU Crossmatch table. The Gaia ephemeris (Section 3.2.3), and the solar-system ephemerides (Section 3.2.1) are also needed.

The IGSL (Initial Gaia Source List), described in Smart and Nicastro (2014) is also used to obtain information on the magnitude $G_{\rm RVS}$. Each source observed by the RVS has an associated external $G_{\rm RVS}$ which is the IGSL magnitude for the large majority of stars treated in Gaia DR2. It is the on-board $G_{\rm RVS}$ for the $\sim 8$% of the sources not belonging to the IGSL.

## Filters applied

This pipeline version treats unblended spectra of bright stars ($G_{\rm RVS}$ $\leq 12$). The majority of the spectra observed by RVS (being fainter) do not satisfy these requirements and are filtered out. They will be processed in future data releases. The following filters are applied to the input data:

• The spectra of the sources having external $G_{\rm RVS}$ $>12$ are filtered out. This selection excludes more than 95 % of the spectra acquired by RVS (the faintest RVS spectra can reach on-board magnitude $G_{\rm RVS}$ $=16.2$).

• The spectra having non-rectangular truncated windows, because in conflict with other source’s spectra, are filtered out. This filter is at transit level: depending on the particular observation geometry during a transit, a source spectrum can overlap or not with one or more other source’s spectra. This filter excludes about 40 % of the spectra. The overlapping is more important in crowded regions, where 60–80 % of the spectra have a truncated window.

• The sources not having AGIS coordinates are removed from the chain (the good astrometric positions provided by AGIS are indispensable for wavelength calibration and thus for the radial velocity estimation).

• The spectra acquired during pre-defined data intervals, where the data are known to be of bad quality are excluded. These intervals include the time during decontaminations, the refocusing, the commissioning of a new version of the on-board VPU (Section 1.1.3), the time intervals where the AGIS residuals are high and the time gaps in the IDT BaryVeloCorr larger than 10 minutes. The total time covered by all the intervals excluded is about 200 revolutions: 7.5 % of the total observation time.